From: "jason pollard" <jasonpollard@yahoo.com>
> Any of you java blackbelt gurus (and you know who you are) have an opinion
> on how/if it would be possible to dynamically load and unload i-appli
canvas
> or panel classes in an effort to circumvent the 10K appli size limit?
>
> I'm thinking maybe one could download some bytecode as a string, cast it
as
> a panel or canvas class, and set it as the current screen.
I'm definitely whitebelt and wannabe, but this strikes me as the kind
of security hole you could fly a B52 bomber through. An iAppli seems
like a virtual ROM cartridge to me. Very much by design.
The only way around this that I can see is to write an iAppli that has
its own VM for some kind of scripting language, armed with lots of
the more likely API bindings.
Which is ... vaguely plausible. Time was, there was this disastrous cult
language called FORTH -- hey don't all you old-timers have an LSD
flashback at once, OK? It's now written "Forth", in one of its few
concessions to civilization.
Some diligent hackers squeezed Forth 'nuclei' (as the cognoscenti
called them) into 4K, maybe less.
Well, Forth is still around. There are Java implementations of it, though
nobody has succeeded in doing a Java-to-Forth-instructions compiler, to
my knowledge, despite a rough philosophical alignment in their VM
architecture (Forth's computation model, like the Java VM, is stack-
based.)
Anyway, if you wanted to have an iAppli that kept fetching code from
the server, doing (*gack*) code overlays, etc., I think, for the time
being, you're stuck with some such approach as this. Programming in 5K
seems, by report, like trying to fill a demitasse cup from a firehose, so if
you could pull it off, it might even be a popular approach to extending
application size. Imagine that -- horrible little stack languages, popular
again! Be afraid, be very afraid. Scarier still: RPN syntax is,
grammatically,
a better match to Nihongo's Subject-Object-Verb word order, so the
usual initial gaijin-programmer objection to Forth ("who wants to write
code backwards?") doesn't really apply.
Of course, if you could do this, it would leave only about ... oh, maybe
12 bytes for actual program space. Some people might be disappointed.
So don't tell anybody I suggested this.
-michael turner
leap@gol.com
Received on Thu Feb 7 06:53:50 2002